Feature

Multi-LLM Support

Choose the AI model that fits your needs and budget

BPract Agents is not locked to a single AI provider. The platform supports OpenAI, Anthropic Claude, OpenRouter, and custom API-compatible endpoints. You can bring your own API key for full control over model selection and costs, or use the platform-provided key for a zero-configuration setup. Switch models per tenant without changing any code.

Key Benefits

Why Multi-LLM Support matters for your business.

1

Native support for OpenAI GPT models, Anthropic Claude models, and any OpenRouter-accessible model

2

Bring your own API key for full control over billing, rate limits, and model selection

3

Platform API key option included in paid plans for zero-configuration AI access

4

Switch models per tenant from the admin panel without touching code or redeploying

5

Custom endpoint support lets you connect to self-hosted or fine-tuned models via any OpenAI-compatible API

Supported Providers

  • OpenAI: GPT-4o, GPT-4o mini, GPT-4 Turbo, and all current chat completion models
  • Anthropic: Claude Haiku, Claude Sonnet, and Claude Opus via the native Anthropic SDK
  • OpenRouter: Access to hundreds of models from Mistral, Meta Llama, Google Gemini, Cohere, and more through a single API
  • Custom: Any endpoint that implements the OpenAI chat completions API spec, including vLLM, Ollama, and Azure OpenAI

Bring Your Own Key vs Platform Key

BPract Agents gives you two options for AI access. With bring-your-own-key, you enter your OpenAI, Anthropic, or OpenRouter API key in the admin panel and all AI usage is billed directly to your provider account. This gives you full control over model selection, rate limits, and costs. Alternatively, paid plans include a platform API key with a daily token budget, so you can get started without setting up any provider accounts. The platform key uses Claude Haiku for optimal cost-efficiency.

Per-Tenant Model Configuration

In a multi-tenant setup, each tenant can use a different AI provider and model. Your marketing site chatbot might run on GPT-4o for creative responses, while your technical support agent runs on Claude Haiku for cost-efficient factual answers. Model changes take effect immediately with no downtime and no redeployment required.

Frequently Asked Questions

Common questions about Multi-LLM Support.

Which AI model does BPract Agents use by default?
The default model is Claude Haiku 4.5, chosen for its excellent balance of quality and cost at roughly one cent per conversation. You can change the model at any time in the admin panel.
Can I use a fine-tuned model with BPract Agents?
Yes. If your fine-tuned model is accessible via an OpenAI-compatible API endpoint, you can configure it as a custom provider in the admin panel. This works with self-hosted models running on vLLM, Ollama, or similar inference servers.
Is there a token budget or usage limit?
When using the platform API key, each tenant has a configurable daily token budget (default 500K tokens per day) to prevent runaway costs. When using your own API key, there is no platform-side token limit since billing goes directly to your provider.

Related Features

Explore more capabilities of BPract Agents.

RAG Knowledge Base

Ground your AI agent in your actual business knowledge

Enterprise Security

67 security controls protecting your AI agent and your visitors

Visual Admin Panel

Configure everything without writing a single line of code

Experience Multi-LLM Support

See how Multi-LLM Support can transform your website. Start your free trial today.

Start Free TrialTalk to Sales